The last decade has seen an increase in the application of machine learning (ML) methods to PET/MRI attenuation correction (AC). This systematic review provides a head-to-head comparison between state-of-the-art ML methods and clinical standards for AC to determine the clinical feasibility of ML approaches PET AC. We extracted numerical values for image quality, tissue classification, regional and global diagnostic performance. The pooled mean relative error for global performance was 0.87 ± 1.3%, the quality of evidence for all outcomes ranged from moderate to very low. Our findings suggest that ML-AC performance is within acceptable limits for clinical PET/MR neuroimaging.
This abstract and the presentation materials are available to members only; a login is required.